* Paging out sleepy processes?
@ 1999-10-20 8:24 Jeff Garzik
1999-10-21 5:38 ` Wang Yong
0 siblings, 1 reply; 3+ messages in thread
From: Jeff Garzik @ 1999-10-20 8:24 UTC (permalink / raw)
To: linux-mm
I have a simple app that I run locally that allocates and randomly
dirties a lot of memory all at once, with the intention of forcing Linux
to swap out processes.
How possible/reasonable would it be to add a feature which will swap out
processes that have been asleep for a long time?
IMHO this behavior would default to off, but can be enabled by
specifying the age at which the system should attempt to swap out
processes:
# tell kernel to swap out processes which have been asleep
# longer than N seconds
echo 7200 > /proc/sys/vm/min_sleepy_swap
Is there a way to do this already?
Regards,
Jeff
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org. For more info on Linux MM,
see: http://humbolt.geo.uu.nl/Linux-MM/
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: Paging out sleepy processes?
1999-10-20 8:24 Paging out sleepy processes? Jeff Garzik
@ 1999-10-21 5:38 ` Wang Yong
1999-10-21 8:09 ` Jeff Garzik
0 siblings, 1 reply; 3+ messages in thread
From: Wang Yong @ 1999-10-21 5:38 UTC (permalink / raw)
To: Jeff Garzik; +Cc: mail list linux-mm mail list
why do u want to force it out. in my opinion, mm need to swap some process
out when the amount of physical
memory is under some limit.
MM will force it out if MM choose this process to swap out.if MM not choose
it, that's to say free memory is
enough or some other processes should be swapped out.
Jeff Garzik wrote:
> I have a simple app that I run locally that allocates and randomly
> dirties a lot of memory all at once, with the intention of forcing Linux
> to swap out processes.
>
> How possible/reasonable would it be to add a feature which will swap out
> processes that have been asleep for a long time?
>
> IMHO this behavior would default to off, but can be enabled by
> specifying the age at which the system should attempt to swap out
> processes:
>
> # tell kernel to swap out processes which have been asleep
> # longer than N seconds
> echo 7200 > /proc/sys/vm/min_sleepy_swap
>
> Is there a way to do this already?
>
> Regards,
>
> Jeff
> --
> To unsubscribe, send a message with 'unsubscribe linux-mm' in
> the body to majordomo@kvack.org. For more info on Linux MM,
> see: http://humbolt.geo.uu.nl/Linux-MM/
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org. For more info on Linux MM,
see: http://humbolt.geo.uu.nl/Linux-MM/
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: Paging out sleepy processes?
1999-10-21 5:38 ` Wang Yong
@ 1999-10-21 8:09 ` Jeff Garzik
0 siblings, 0 replies; 3+ messages in thread
From: Jeff Garzik @ 1999-10-21 8:09 UTC (permalink / raw)
To: wung_y; +Cc: mail list linux-mm mail list
Wang Yong wrote:
> Jeff Garzik wrote:
> > How possible/reasonable would it be to add a feature which will swap out
> > processes that have been asleep for a long time?
> why do u want to force it out
Processes will not get swapped out until memory pressure occurs. Thus,
idle processes waste physical memory until this situation occurs.
Physical memory is always a valuable commodity, and should IMHO be
reclaimed whenever possible.
It makes even more sense to page out idle _pages_ instead of processes,
but I don't know if it is even possible to determine how long a page has
been idle, in terms of clock time.
Jeff
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org. For more info on Linux MM,
see: http://humbolt.geo.uu.nl/Linux-MM/
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~1999-10-21 8:09 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
1999-10-20 8:24 Paging out sleepy processes? Jeff Garzik
1999-10-21 5:38 ` Wang Yong
1999-10-21 8:09 ` Jeff Garzik
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox